Deep Learning Model for Pigeon Pea Leaf Disease Identification

The world's main grain crop is pigeon pea. It sustains the mainstream of the world's population with nourishment and liveliness. Various pigeon pea leaf diseases which may occur by fungal, bacterial and viral infections and are major threat to the world's agricultural production because they deteriorate crop quality in addition to lowering crop yield. Traditional strategies of classifying and managing disorders frequently rely on laborious and subjective manual examination procedures. However, recent advancements in computer vision methods have made it possible to construct automated systems that can recognize illnesses from picture data. In this work major leaves spoiled by disorders are classified: leaf spot, sterilic mosaic and leaf webber along with healthy pigeon pea plant leaves. The usefulness of the Convolutional Block Attention Module (CBAM) and squeeze-and-excitation (SE) attention processes are examined in this study. Renowned CNN models, such as ResNet50 and VGG19, are used to perform transfer learning for the categorization of plant diseases and are subsequently improved. The results of this study demonstrate that ResNet50 + CBAM stands out as the superior performer, showcasing remarkable metrics across various evaluation criteria. Interestingly, ResNet attains a noteworthy 0.97 accuracy.